“The OpenAI nonprofit board has the ultimate authority to decide whether or not to deploy a new model, and can halt a deployment for safety reasons even if there is a strong commercial incentive.”